Economic Model

An economic model is a simplified description of how some part of the economy works. It picks out a few key factors, such as prices, incomes, interest rates or expectations, states how they are related, and leaves out everything else. Models can be verbal arguments, diagrams like the familiar supply and demand cross, or systems of equations solved by computer. Economists use them to explain behaviour, to test whether a theory fits the data and to predict the effects of policies, from a tax change to a rise in the federal funds rate.

Models are the basic working tool of both microeconomics and macroeconomics. They are estimated and tested with econometrics, and their predictions are checked against evidence from natural experiments and trials. The statistician George Box's remark that "all models are wrong, but some are useful" sums up the standard view: a model is judged by whether it helps answer a particular question, not by whether it captures reality in full. Most models start from scarcity, the condition that limited resources cannot meet every want, and then describe how people, firms or governments choose between alternatives.

A road map is a model of a country. It leaves out trees, houses and the colour of the cars, and that is exactly why it is useful: it shows the roads you need to get from one town to another. A map that included every detail would be as big and confusing as the country itself.

An economic model works like that map. To ask what happens to bread sales when flour gets more expensive, it ignores most of the economy and keeps only bakers, buyers and prices. The art lies in leaving out what does not matter for the question while keeping what does. A map for hikers and a map for drivers look different, and so do models built for different questions.

A classic textbook model of national income has households spending a fixed amount plus a fraction of their income, and total output equal to spending:

• $Y$ is national income (output).
• $C_0$ is spending that does not depend on income.
• $c$ is the marginal propensity to consume, the share of each extra unit of income that is spent.
• $I$ is business investment and $G$ is government spending.

Solving gives $Y = (C_0 + I + G) / (1 - c)$. With $C_0 = 100$, $I = 200$, $G = 200$ and $c = 0.8$, income is $500 / 0.2 = 2{,}500$. If government spending rises by 10, income rises by $10 / 0.2 = 50$: a multiplier of 5. Real economies have taxes, imports, interest rates and prices that respond, so actual multipliers are much smaller, which illustrates how a model's assumptions drive its answers.

• Theoretical models. Stylised models, such as game theory models of strategic behaviour or the Phillips curve relation between unemployment and inflation, isolate a mechanism.
• Equilibrium models. General equilibrium models trace how changes in one market spread to others, building on the idea of economic equilibrium.
• Macroeconometric models. Large systems of estimated equations, such as the Federal Reserve Board's FRB/US model, used for forecasting and policy analysis.
• DSGE models. Dynamic stochastic general equilibrium models, built on households and firms making forward-looking choices with rational expectations, are used by central banks including the ECB and the Bank of England.
• Agent-based and statistical models. Simulations of many interacting agents, and data-driven forecasting models, including machine learning, complement the traditional approaches.

Early economists reasoned with verbal models; David Ricardo's arguments about trade and rent are often seen as an early example of abstract modelling. In 1949 the New Zealand-born economist Bill Phillips built the MONIAC, a hydraulic machine that modelled the flow of money through the British economy with coloured water. Jan Tinbergen constructed some of the first statistical models of national economies in the 1930s, work recognised with the first Nobel Memorial Prize in Economic Sciences in 1969. After Robert Lucas's 1976 critique, which argued that models estimated on past behaviour can fail when policy changes, research shifted towards models built on explicit choices and expectations.

Central banks such as the Federal Reserve use suites of models to forecast inflation and growth and to estimate quantities that cannot be observed directly, such as the output gap and the neutral interest rate. Finance ministries use them to cost tax and spending plans; businesses and banks use them for demand forecasting and stress testing.

Criticism has come from several directions. After the 2008 financial crisis, many mainstream macroeconomic models were faulted for leaving out banks, leverage and financial panics. Behavioural economists question assumptions of fully rational agents, and heterodox economists, including followers of the Austrian school and post-Keynesians, argue that formal models can give a false sense of precision, a charge sometimes summed up as physics envy. Defenders reply that models make assumptions explicit and testable, so their errors can be found and fixed, and that models have since been extended to include financial frictions and varied household behaviour.

• Federal Reserve Board: FRB/US model
• Bank of England: The Bank's forecasting platform (COMPASS)